The Limits on Sexual Dimorphism in Vegetative Traits in a Gynodioecious Plant Author(S): Tia-Lynn Ashman Reviewed Work(S): Source: the American Naturalist, Vol

Home , Dioecy, Jojoba, Rhus typhina

The Limits on Sexual Dimorphism in Vegetative Traits in a Gynodioecious Plant Author(s): Tia-Lynn Ashman Reviewed work(s): Source: The American Naturalist, Vol. 166, No. 4, Sexual Dimorphism: Its Causes and Consequences (Oct., 2005), pp. S5-S16 Published by: The University of Chicago Press for The American Society of Naturalists Stable URL: http://www.jstor.org/stable/3473063 . Accessed: 06/04/2012 11:34

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press and The American Society of Naturalists are collaborating with JSTOR to digitize, preserve and extend access to The American Naturalist.

http://www.jstor.org VOL. 166, SUPPLEMENT THE AMERICAN NATURALIST OCTOBER 2005

The Limitson SexualDimorphism in VegetativeTraits in a Gynodioecious Plant

Tia-LynnAshman*

DepartmentBiological Sciences, University of Pittsburgh, often not as pronouncedor absentaltogether. For example, of Pittsburgh,Pennsylvania 15260; and Pymatuning Laboratory leaf size has been found to be sexuallydimorphic, at least 16424 Ecology,Linesville, Pennsylvania in some sites or under some environments, in all (26 of 26) dioecious species studied (table 1), but it is not dimorphic in the majority (five of six) of gynodioecious species studied (table 1). Leaf, stem, and branch number ABSTRACT: Gynodioecious plants exhibit modest sexual dimorphism (i.e., degree of ramification) are also more likely to be in vegetative and phenological traits, which stands in stark contrast dimorphic in dioecious (25 of 26) than in gynodioecious to in traits. I evaluate the roles pronounced dimorphism reproductive (three of four) species (table 1). Such patternsmight sug- of limited genetic variation, negative genetic covariation (within and that dimorphism in vegetativetraits evolves after the between sex morphs), and lack of gender-differential selection in gest evolution of in to sex-differ- contributing to minimal sexual dimorphism for these traits in Fra- separate genders response garia virginiana. Major findings are as follows. First, selection was ential selection (Geber 1999), but few have explored this sometimes differential but rarely divergent between male and female evolutionarymechanism (but see Kohorn 1994;Bond and fertility modes. Second, response to selection was constrained by low Maze 1999). In fact, we currentlyhave little data that can variation and extensive covariance. In covariance genetic genetic fact, address the question, why are vegetative traits in gyno- between traits within sex morphs appears to represent a constraint dioecious species not more dimorphic? on with that of covariance between sex morphs. Third, these par Lack of could result from absent/variable constraints combine with different modes of gamete transmission to dimorphism produce very different gender-specific contributions to the mean selection for dimorphism or from genetic constraintslim- phenotypes of the next generation. Finally, predicted responses to iting a response to divergent selection. It has been hy- selection for several traits are concordant with the degree and di- pothesized that life-history and vegetative traits ought to rection of in a related dioecious In dimorphism closely species. sum, be under divergentselection through male and female fer- this work suggests that minimal sexual dimorphism in vegetative and tility (e.g., S,, - S,) because of differential costs or re- phenological traits is due to similar directional selection via male of these modes of 1999; and female fertility combined with the constraints of low genetic quirements reproduction (Delph variation and extensive genetic covariance both within and between Geber 1999;Case and Ashman2005). For instance,females sex morphs. may be selected to flower later (or less often), invest less in asexualreproduction, or grow more than males,because adaptive evolution, antagonistic covariation, dioecy, Fra- Keywords: more total resources are required to successfullymature garia virginiana, G matrix, natural selection. fruit than to produce pollen. Alternatively,because allocation to male function incurs greater opportunity costs allocationto "tradesoff'" with In many plants, sexual dimorphism extends beyond pri- during flowering(i.e., pollen allocationto Eckhartand mary sexual traits into life-history and vegetative traits vegetativegrowth; Chapin 1997), males be selected to fewer leaves and (Dawson and Geber 1999; Delph 1999), but this is not may produce grow less than females. It is not how- universal.In fact, while gynodioeciousspecies (femalesand during flowering known, whether in dioecious hermaphrodites)often show sexual dimorphism in repro- ever, vegetative dimorphism species for ductive traits (e.g., flower size and number) that rivalsthat representsdirect responsesto the distinct requirements seen in dioecious species (Eckhart 1999; Shykoff et al. male and female reproduction (Kohorn 1994; Machon et 2003), the degree of dimorphism in vegetative traits is al. 1995), indirect responses to selection on reproductive characters(Bond and Midgley 1988), or both. For instance, * E-mail:[email protected]. leaf size could be the direct target of selection via female function if leavesbetter Am. Nat. 2005. Vol. 166, pp. S5-S 16. ? 2005 by The Universityof Chicago. larger support adjacentdeveloping 0003-0147/2005/1660S4-40864$15.00.All rightsreserved. fruits. In contrast, leaf size in males could be selected S6 The American Naturalist

Table 1: Studies of sexual dimorphism in vegetativetraits in dioecious, subdioecious, and gynodioecious species Sexual Branch or stem Species system Leaf size Leaf number number Reference Asparagusofficinalis D F > M F < M Dzhaparidze1967; Machon et al. 1995 Cannabissativa D F > M F > M Dzhaparidze1967 Diospyrosvirginiana D F > M Dzhaparidze1967 chiloensis D F M Hancock and Bringhurst1980 Fragaria < Leucadendronspp. (17 species) D F> M F < M Bond and Midgley 1988 (16 of 17 spp.) Leucadendronxanthoconus D F > M F < M Bond and Maze 1999 Linderabenzoin D F > M F M F < M Dzhaparidze1967 Populusgradidentata D Sakai and Sharik 1988 Populus tremuloides D Sakai and Burris 1985 Rhus typhina D F < M F M Korpelainen1992 Rumex acetosella D F < M, F > M F > M Lovett Doust et al. 1987 Silene latifolia D F > M F = M F > M Meagher 1992; Lyons et al. 1994 Simondsiachinensis D F > M Kohorn 1995 Siparunagrandiflora D F < M F > M F > M Nicotra 1999 Cucurbitafoetidissima G F = H F = H Kohn 1989 Daphne laureola G F = H Alonso and Herrera2001 Lobeliasiphilitica G F < H Caruso et al. 2003 Plantagolanceolata G F = H F > H Olff et al. 1989; Poot 1997 Sidalceaoregana spicata G F = H F < H T.-L. Ashman, unpublished data Phacelia linearis G F> H Eckhartand Chapin 1997 Wurmbeabiglandulosa G F = H Ramsey and Vaughton 2001 Wurmbeadioica SD F = H = M Ramsey and Vaughton 2002 Note: Direction(e.g., F > M, F < M) of statisticallysignificant sexual dimorphism is noted for each trait and speciesstudied. If sexualdimorphism was nonsignificantin the originalstudy, it is denoted as F = M. D = dioecious;SD = subdioecious;G = gynodioecious;F = female;H = hermaphrodite; M = male. indirectlyas a consequence of selection for increasedpol- than one sexual morph because the response to selection len production (or dispersal)if pollen production is pos- is a function of three forms of genetic covariation:between itively correlatedwith branch number,which is negatively traits within the sexes, between homologous traits across correlated with leaf size (Bond and Midgley 1988). Al- the sexes, and between nonhomologous traits across the though studies that quantify selection on vegetativetraits sexes (Morgan and Ashman 2003). These ideas, however, in gender-dimorphicspecies have been repeatedlycalled have not been formally explored in a gynodioecious for, few have been conducted (but see Kohorn 1994;Bond species. and Maze 1999), and none have used selection gradient Both positive and negativegenetic correlationscan con- analysis(Lande and Arnold 1983) as a means of describing strain adaptive evolution. First, positive genetic correla- and assessingvariation in selection. tions can slow evolution if correlatedtraits are selected in If there is divergent selection on a trait via male and opposite directions (Conner and Via 1992). Second, neg- female fertility but sexual dimorphism is not seen, then ative genetic correlations can slow adaptive response if low genetic variation or the presence of genetic correla- selection favors high values of both traits. In fact, this sort tions may be limiting response to selection. Vegetative of "antagonistic"correlation can reversea responseto se- traits have been shown to harbor high levels of genetic lection if one trait is under positive selection while the variationin hermaphroditeplants, leading Geber and Grif- other is under negative selection. Size-numbertrade-offs fen (2003) to suggest that lack of genetic variation may are a pervasive example of antagonistic covariation, and not be the most common constraint to adaptive evolu- these have received much attention at the level of repro- tionary change in these plants. However, genetic corre- ductive units (e.g., flowers;Worley and Barrett2000; Ca- lations may be particularlyimportant in specieswith more ruso 2004; Delph et al. 2004), but they also occur for Limits on SexualDimorphism S7 vegetative modules (e.g., leaves, branches) and can con- sexuallyvia seed or vegetativelyvia stolons (runners).Both strain evolutionaryresponse in these traits as well (Etter- inflorescences and runners develop from axillary buds. son and Shaw 2001). Third, correlationsbetween traits in Females have very high fruit set (~90%), while that of the sex morphs can result in correlatedresponses in one hermaphroditesis low (~20%), even under optimal con- morph that are greaterthan, less than, or the opposite of ditions (Ashman 2003). Wild populations vary in the fre- what is predicted from selection in that morph alone quency of females, and females produce the majority of (Meagher 1999; Morgan and Ashman 2003). Few studies the seed in populations in northwestern Pennsylvania have determined the degree to which genetic correlations (Ashman 1999). alter selection response (but see Campbell 1996; Mitchell et al. 1998;Etterson and Shaw 2001; Caruso2004), despite Plant Material, Cultivation,and Scoring overwhelming evidence that direct selection accounts for only part of total selection, especiallyon functional traits The intent of this study was to estimate selection, genetic in plants (Geber and Griffen 2003). (co)variances, and predicted response to selection for a To investigatethe microevolutionarydynamics of sexual synthetic field population of F. virginiana,so I used plants dimorphism, I conducted a phenotypic selection analysis from a nested breeding design that were grown in a field of several vegetative traits and a phenological trait in a garden. Here, I briefly summarize cultivationprocedures, synthetic field population of gynodioecious wild straw- while full details are available in Ashman (2003). Plant berry (Fragariavirginiana). I also estimated genetic vari- material was originally obtained from three wild popu- ation and covariation for these traits under field condi- lations (PR, HT, W) in Crawford County, Pennsylvania tions, providing the most appropriatemeasures to predict (see Ashman 1999 for details). Within each population, rate of adaptive evolutionary change (Geber and Griffen crosses were performedto yield 210 maternalfull-sib fam- 2003). Information on selection and genetic (co)variation ilies nested within 70 paternal half-sib families. In two were combined to predictthe multivariateadaptive change replicate plantings (2000 and 2001), seedlings from these across one generation (Lande and Arnold 1983; Morgan crosses were planted in a completely randomized design and Ashman 2003). I evaluatedwhether the traits would in a field garden adjacentto the Pymatuning Laboratory be expected to evolve in a way that reflected selection, of Ecology, CrawfordCounty, Pennsylvania.Plants were whether genetic architecturewould constrain a response protected from deer and goose herbivorywith netting ex- to selection, and whetherthe responsewas in the direction closures;however, no other attempts were made to retard of greater sexual dimorphism. In doing this, I sought to other naturalherbivores or recolonizationof naturalveg- answer the following questions. First, what is the pattern etation afterplanting. Insect pollination was supplemented of phenotypic selection between male and female fertility by hand pollination three times per week. The frequency or between sex morphs (female fertility only)? Second, of females among flowering plants was 50% and 55% in what is the level of genetic variation and covariation 2000 and 2001, respectively. (within and between sex morphs) for vegetativeand phe- For up to eight floweringplants per maternalfamily (2 nological traits?And third, does genetic architecturelimit progeny/sexmorph/year: a total of 1,440 in 2000 and 2,011 response to selection? In particular,does genetic covari- in 2001), I recorded leaf size and number at the time of ation facilitate,retard, or reverseresponse to selection, and transplanting (hereafter "fall") and again at the end of do genetic covarianceswithin or between the sex morphs flowering (hereafter"spring"). At each measurementtime, have the most important influence on the response to leaf size was estimatedon one fully expandedleaf per plant selection? that was ~1 month old. The width of the central leaflet was measured on the fall leaves, whereas the "best- fit" ellipse (National Institutes of Health Image program; Methods http://rsb.info.nih.gov/nih-image)of the areaof the central leafletwas measuredon the leaves.These two meth- Study Species spring ods of estimating leaf size are strongly correlated (r = Fragariavirginiana (Rosaceae), the Virginian wild straw- 0.998, N = 20, P< .0001). I also recordedthe number of berry,is a creepingperennial herb that is native to eastern trichomes on the upper leaf surface of three 154 mm2 North America (Staudt 1989) and commonly grows in circularsections from spring leaves (the averageof these meadows, in old fields, and along road or forest edges. It was used in analyses),the date of first flower (Juliandate), has a gynodioecious sexual system where females and her- and the number of runnersproduced per plant by the end maphrodites coexist. Sex determination is under nuclear of flowering.As part of a previous study, total flower and control, with male sterility(femaleness) dominant to male fruit production, pollen per flower, and ovules per flower fertility (Ahmadi and Bringhurst1989). Plants reproduce had been recorded (for details, see Ashman 2003), and S8 The AmericanNaturalist they were used here to estimate female and male fertility the methods describedin Lynchand Walsh (1998). I esti- (see "Phenotypic Selection"). MANOVA and univariate mated additive genetic varianceas four times the sire var- ANOVAs (with year and sex as fixed effects and popula- iance component and calculatedthe narrow-senseherita- tion, sire, and dam as nested random effects) were per- bility as the additivegenetic variancedivided by the total formed to evaluate the sources of trait variation and, in phenotypic variance. I determined significanceof herita- particular,to determine whether there were differences bilities by a significantsire effect in the ANOVA and by between the sex morphs in vegetative and phenological randomizationtests, as in Ashman (2003). In all cases,ran- traits. domization tests agreedwith ANOVAresults, so only the ANOVAresults are presented.I compared h2 between sex morphs, using t-tests with bootstrappedstandard errors PhenotypicSelection (e.g., Roff 1997;Ashman 2003) followedby Bonferroniad- Differentforms of selection can operatein each sex morph justmentsfor multiple tests. and through male and female fertilitiesof hermaphrodites I calculated genetic correlations and covariances be- (Morgan and Ashman 2003), so I estimated standardized tween traits within each sex morph and within traits be- phenotypic selection gradients via female fertility of fe- tween the two sex morphs, using the family-structured males (hereafter,females' fertility;I f'), female fertility of progeny data. I performedPearson product-moment cor- hermaphrodites(f3f), and male fertilityof hermaphrodites relations on paternalhalf-sib family means that were ob- (#'Hm) in each year from multivariateregressions of relative tained by first averagingsame-sexed progeny from each fertilitieson standardizedtrait values. Femalefertility was dam and then averagingover all three dams per sire. This estimated as the product of fruit production and ovules method was used to similarly produce estimated covari- per fruit. This is a good estimate of total seed production ances within and between the sex morphs, but it may be because seed set per ovule is high under hand pollination subject to some bias (Lynch and Walsh 1998). To deter- (Ashman and Diefenderfer2001). Male fertility was esti- mine whether correlationsbetween homologous traits in mated by assigning each hermaphroditea portion of the the sex morphs were different from 1, I performed one- seeds produced within a year based on its relativepollen tailed t-tests,using bootstrappedstandard errors (e.g., Roff production. I calculatedpollen production as the product 1997). I compared correlations between sex morphs as of total flower number and pollen per flower. I evaluated above. the interaction term between year and traits in ANCOVA I tested for overall sex differencesin the G matrix (i.e., to determine whether gradients differed between years. deviations from matrix equality G, = GH) by calculating Becauseyear-to-year variation was very limited (only two Roffs T statistic and using randomization to determine of 21 selection gradients showed significantyearly varia- significance(B&gin and Roff 2001). I also computed Roffs tion [runner number #' = -0.0486 vs. 0.0844; #'m = T%/o,which indicates the overall absolute difference be- 0.0088 vs. 0.1974]), I calculatedstandardized linear selec- tween elements in the two matricesas a percentageof the tion gradients, using data pooled across years after the size of the matrix elements. effect of year was removed using ANOVA. I present only these estimates and use them to predict response to selection. Because all plants were hand-pollinatedand male Estimationof PredictedResponse to Selectionand estimates are based on pollen production alone, fertility Evaluationof Constraints traits under selection should be mediated by resourceac- quisition and allocation (i.e., to meet the differing costs Total multivariate response to selection AZH) and of male and female function) rather than by aspects of gender-specificcomponents of total response(AF,, to selection pollination. (GFF,, GHF"3Hm GHF,,F and GH3Hf) were estimated from the phenotypic selection gradientspooled acrossyears and Genetic(Co)Variation, Heritabilities, and the genetic (co)variance matrices according to the for- GeneticCorrelations mulas developed by Morgan and Ashman (2003), that is,

Becausethe intent of this study was to estimate response 1 to selectionfor a syntheticpopulation under field conditions AZF -= (GFIF + GHFIHm), (1) ratherthan to assess population variabilityin genetic variation, I statisticallyremoved the effects of both planting year and population from the data, using ANOVA,before 1 estimatinggenetic (co)variances.I estimatedvariance com- + AZH = [(GHFaIF + GHbjHf) GHjHm]. (2) ponents and heritabilitiesfrom nested ANOVA,following 2 Limits on Sexual Dimorphism S9

Here GF, GH, and GHF are the genetic (co)variancematrices Table2: Phenotypicmeans (SE) for vegetativeand phenolog- within females, within hermaphrodites,and between the ical traitsof femaleand hermaphroditeFragaria virginiana in a field sex morphs, respectively, whereas #,,F Hf, and #Hm are the grown garden selection gradientsvia females' fertility,female fertility of Trait Female Hermaphrodite and male of re- hermaphrodites, fertility hermaphrodites, Flower date (Juliandate) 116.80 (.104) 116.40 The a and b reflect (.112) spectively. parameters fitness-weighted Fall leaf number 8.76 (.048) 8.93 (.047) sex ratio (Morgan and Ashman 2003). Unstandardized Fallleaf size (mm2) 32.41 (.094) 32.08 (.091) selection gradients were used to determine the gender- Spring leaf number 19.56 (.146) 19.95 (.147) specific components of the response to selection. All es- Spring leaf size (mm2) 21.17 (.209) 20.96 (.195) timates of selection were used in calculatingthe responses Trichomenumber 7.90 (.129) 7.79 (.129) to selection regardlessof significance because these rep- Runnernumber 8.36 (.100) 8.89 (.104) resent the best unbiased point estimates of selection. Note: Means in boldface are significantlydifferent between the sex To evaluate whether a response to selection is con- morphs,as determinedby a significantsex effectin a mixed-modelnested strainedby genetic (co)variation,I calculatedstandardized ANOVA.N = 3941-3986. gender-specificcontributions to responseto selection (e.g., the unstandardized the and dams, as well as environmentallyinduced variation GHF/•) by dividing responses by phenotypic standard deviation for that sex morph and (years) for all traits (data not shown). compared these with the standardizedselection gradients t-tests followed Bonferroni If a stan- by by adjustments. PhenotypicSelection dardized component of response to selection (e.g., GHF,,) is smaller in absolute value or differs in sign from Across years, selection was generallyweak; however, some the standardized selection gradient (e.g., #'), then this gradientswere strong, and common patternsemerged (ta- suggests genetic constrainton responseto selection (Con- ble 3). First, there was significant selection through fe- ner and Via 1992). To further explore the source of con- males' fertility to increase spring leaf number (3#' = straint, I partitioned gender-specificcontributions to se- 0.0640). Second, selection always favored flowering early, lection response into those due to direct and indirect and this was strongest through female fertility of her- responses (i.e., due to selection on correlatedtraits), fol- maphrodites(#', = -0.2111). Third, trichomeswere un- lowing Conner and Via (1992). If the indirect response to der selection to decrease via all fertility modes, but this selection is in the opposite direction of the directresponse did not reach statisticalsignificance. Fourth, fall leaf num- to selection, then this suggeststhat genetic correlationsare ber and size were always under selection to increase,sig- involved in constrainingthe evolution of the trait (Conner nificantly so via females' fertility and male fertility. and Via 1992). I used estimates of contributions to re- gender-specific Genetic(Co)Variation, Heritabilities, and to selection- and sex ratio sponse fitness-weighted param- GeneticCorrelations eters in equations (1) and (2) to predict the standardized between-generationchange of each trait in females and Heritabilities ranged from 0.108 to 0.542, and all were hermaphrodites.From these values I predicted the con- statisticallydifferent from 0 but not differentbetween the sequences for sexual dimorphism. sex morphs (fig. 1). Similarpatterns of genetic correlation were seen within the sex morphs (fig. 1), and none differed significantlybetween the sexes afterBonferroni adjustment Results for multiple tests. Strong positive correlationsexisted between runner number and spring leaf number (0.715 and StandingSexual Dimorphismand PhenotypicVariation 0.691 for hermaphrodites [H] and females [F], respectively), between fall and spring leaf size (0.396 [H], 0.251 Although only a small degree of sexual dimorphism was [F]) and number (0.449 [H], 0.397 [F]). Trade-offswere found for vegetative traits and flowering time (table 2), evident for leaf size and number in both seasons and sex the overall sex effect was significant (MANOVA: F = morphs (fall: -0.244 [H], -0.142 [F];spring: -0.137 [H], 4.14, df = 7,3,277, P < .0002), and univariate ANOVAs -0.201 [F]), as well as between leaf traits across seasons indicatedthat three traitsdiffered significantly between the (fall leaf number-spring leaf size: -0.149 [H], -0.233 sexes. Females produced 6% fewer runners and 3% fewer [F]). Significantnegative correlations were also presentfor spring leaves and flowered approximatelyhalf a day later spring leaf size and trichomes in both sexes (-0.316 [H], than hermaphrodites.There was also evidence for signif- -0.353 [F]). Corroboratingthe between-sex tests above, icant genetic variation at the level of populations, sires, a randomizationtest did not detect a differencebetween SI0 The AmericanNaturalist

Table3: standardizedselection and contributions Gender-specific gradients(38') gender-specific (G34',GH.'F, GHFIHm, GHm)im, to the total standardized multivariate to selection GH#Hf) predicted responses (AzF,AIH) to Phenotypic Standardizedgender-specific contributions to predictedresponses selection selection GF~ GHF/'F Trait #'r(SE) Total Direct Indirect Total Direct Indirect Flower date -.0098 (.0155) -.0018 -.0016 -.0002 .0063 -.0038 .0101 Fall leaf number .0496 (.0168)* .0385 .0144 .0241 .0262 .0195 .0067 Fall leaf size .0378 (.0155)* -.0197a .0083 -.0279 -.0060 .0207 -.0267 Spring leaf number .0640 (.0218)* .0463 .0104 .0359 .0163 .0109 .0054 Spring leaf size -.0270 (.0164)** -.0046 -.0028 -.0018 -.0173 -.0067 -.0107 Trichomenumber -.0239 (.0157) -.0152 -.0127 -.0025 -.0014 -.0133 .0119 Runner number .0400 (.0212)** .0373 .0062 .0311 .0064 .0019 .0045

GFOFGFHIHm

/Hm (SE) Total Direct Indirect Total Direct Indirect Flower date -.0417 (.0166)* -.0111 -.0132 .0022 .0028a -.0089 .0117 Fall leaf number .0410 (.0173)* .0098 .0165 -.0067 .0259 .0139 .0120 Fall leaf size .0380 (.0167)* .0091 .0212 -.0121 .0104 .0164 -.0061 Spring leaf number -.0310 (.0246) -.0040 -.0047 .0007 .0762 -.0048 .0810 Spring leaf size .0182 (.0176) -.0006 .0048 -.0054 .0256 .0057 .0199 Trichome number -.0161 (.0171) -.0049 -.0083 .0034 -.0321 -.0050 -.0272 Runner number .1587 (.0238)* -.0003a .0066 -.0069 .0117a .0189 -.0072

DirectGH(SE)Total Indirectf

f'Hf (SE) Total Direct Indirect Flower date -.2111 (.0383)* -.0421a -.0495 .0074 Fall leaf number .0339 (.0406) .0305 .0113 .0191 Fall leaf size .0024 (.0391) -.0063 .0010 -.0073 Spring leaf number -.0038 (.0573) .0531 -.0006 .0538 Spring leaf size .0580 (.0411) .0410 .0176 .0234 Trichome number -.0588 (.0398) -.0579 -.0189 -.0390 Runner number .1292 (.0553)* .0131 .0164 -.0033

Note: Gender-specific contributions are partitioned into the direct response due to selection on the trait and the indirect response via correlated traits. Total responses to selection in boldface are significantly different from the selection gradient by t-test. Significantly different from selection gradients after Bonferroni adjustment for multiple tests. * P < .05. ** .05

0.80

0.40 FD 0.00 FLN .40 S -0 + SLN F? SLS SL 8' T LS TN RN 0.80 0.40 * FD 0.00 FLN FLS -0.40 SLN F SLSLS RN'TN TNRN

Figure 1: Significant (P< .05) heritabilities (diagonal) and genetic correlations (left of the diagonal) between traits within hermaphroditesand females. Parameterssignificant after Bonferroni adjustment for multiple tests are denoted with an asterisk.FD = flowering date; FLN = fall leaf number; FLS = fall leaf size; SLN = spring leaf number; SLS = spring leaf size; TN = trichome number; RN = runner number. and the latteris selectedto decrease(0'3 = -0.0270). Both floweringdate (59%) and trichome number (31%), which of these contribute to the negative indirect response to were under strong selection via this fertility pathway. selection that is about twice the size of the direct response to selection (-0.0279 vs. 0.0083). Inspection of gender- PredictedResponse in Sexual Dimorphism specific contributions to selection response that involve covariation between the sex morphs (e.g., GHF/#F, The change in sexual dimorphism predicted for this set also indicates that this type of covariation re- of traits is illustratedin 3. Flowerdate, fall leaf size, GHFlm) figure stricts direct response to selection much of the time (five and fall leaf number lie close to the diagonal line that of six responses;table 3). reflects no predicted change in sexual dimorphism. This The relativeroles of all gender-specificcontributions to is not surprising,because all were selected similarly (to the total predictedselection response are illustratedin fig- increase or decrease) through both male and female fer- ure 2. A relativelysmall proportion of the predicted re- tilities. The off-diagonal position of the other traits in- sponse in hermaphroditesand females is due to selection dicates predicted changes in sexual dimorphism, albeit occurring in the opposite sex morph (GHFa#F and small ones. Most notable is that spring leaf number is GHFHm, respectively). This can be seen most readily in predictedto increasetwo times as fast in hermaphrodites female response, where the contribution due to selection as in females (0.0510 vs. 0.0211 standarddeviation units) via hermaphrodites'male fertilityis usually less than 30% across one generation.The predicted response in females of the total response.The exception, however,is flowering parallels selection on spring leaf number in females where the of the is due to date, majority (85%) response (f3' = 0.0640), whereas the response in hermaphrodites selection occurring via male fertility of hermaphrodites. reflects the sum of all three positive gender-specificcon- In hermaphrodites,contributions via male fertilityaccount tributions (fig. 2) despite negative selection via male and for 15%(flower date) to 75% (springleaf number) of total female fertility in hermaphrodites.Single-generation pre- response.In addition, despite the fact that hermaphrodites dicted responses also suggest that spring leaf size will in- contribute only 26% of the female gametes to the next crease in hermaphrodites while decreasing in females generationof hermaphrodites(b = 0.26), selection via fe- (0.011 vs. -0.002 standarddeviation units). male fertility of hermaphroditescontributes substantially Despite the fact that predictedchanges reflect small per- to the total response to selection in a few traits, such as centages per generation, if all else is equal (i.e., no op- S12 The AmericanNaturalist

Table 4: Genetic correlations(SE) between the sex morphs of Fragariavirginiana Hermaphroditetrait Spring leaf Trichome Runner Female trait Flower date Fall leaf number Fall leaf size number Spring leaf size number number

Flower date .475 (.048)"a, Fall leaf number .013 (.042) .445 (.045)"a, Fall leaf size .059 (.055) -.112 (.049) .610 (.039)a,* Spring leaf number .086 (.053) .232 (.053)a,* -.180 (.048)* .224 (.052)a,* leaf size Spring -.164 (.047)* -.154 (.050)* .221 (.059)a,* -.307 (.052)a,* .363 (.042)",* Trichomenumber .013 (.067) -.027 (.052) -.014 (.042) .104 (.050) -.143 (.053)* .567 (.037)"a* Runner number .029 (.054) .026 (.052) -.085 (.053) .055 (.054) -.165 (.049)* .135 (.051)** .070 (.058) Note: Correlationsbetween homologous traits are on the diagonal,and those betweennonhomologous traits are off the diagonal. " Parametersthat remainsignificant after Bonferroni correction for multipletests. SP < .05. ** .05

posing selection on other life stages) and selection/genetic in Geber 1999). The only evidence for divergentgender- variances remain stable over time, after 100 generations, specific selection in this study involved springleaf number females are predicted to be 64% less leafy and have 47% and size: selection favored numerous, small leaves via fe- smaller leaves in the spring than hermaphrodites. males' fertility but the reversevia male fertility,although the latter was not statisticallysignificant (table 3). Only one plant study provides data for comparison, and it was Discussion conducted on a sexuallydimorphic dioecious shrub. Spe- Kohorn demonstratedthat different Not unlike many gynodioeciousplants (table 1), the minor cifically, (1994) vegetative benefited male versus female (<6%) degreeof sexual dimorphismin vegetativeand phe- morphologies reproduction in Simmondsiachinensis: small leaves and short nological traits in Fragariavirginiana stands in starkcon- internodeswere associatedwith increasedmale suc- trast to the marked sexual dimorphism in reproductive weakly cess (more while leaves and internodes traits (e.g., petal size and fruit-setting ability; Ashman flowers), large long were associated with female success 2003). The resultsof this study shed some of the first light higher (flower buds, seed size). Kohorn concluded that in on why this might be. First, although strengthand direc- (1994) dimorphism traits was the result of trait tion of selection varied between fertility modes for some vegetative divergent optimal values for male and female success. Unfor- traits, many traits are selected similarly.Second, predicted reproductive limited data con- response to selection is limited not only by low genetic tunately, preclude drawing any general clusions the of variation but also by extensive genetic covariance, and regarding genesis vegetative dimorphism in dioecious or more se- covariance between traits within sex morphs appears to gynodioecious species. Clearly, lection studies are needed to the role of representa constraint on par with that of covariancebe- gradient clarify selection on traits. In tween sex morphs. Third, these constraintscombine with gender-specific vegetative particular, studies of selection on traits the different modes of gamete transmission to produce vegetative through components of male are needed. Because traits very different gender-specific contributions to the next fertility vegetative are often subsumed into a that re- generation of females and hermaphrodites.Finally, these principal component flects overall size Elle and it is elementstogether produce predicted responses to selection plant (e.g., Meagher2000), that are concordant with the direction and level of di- not often possible to determinewhether selection on vegetativetraits differbetween male and female morphism seen in a closely related dioecious species, al- might fertility. Such studies in also would though not with the general pattern seen in dioecious hermaphroditespecies provide much-needed For it is un- species reviewed in table 1. In the following paragraphs, comparativeanalysis. instance, known whether traits in I explore these results in greater detail and consider the vegetative hermaphroditespecies reflect the between selections be- assumptions underlying the predicted changes in sexual compromise conflicting tween male and female as has been for dimorphism. fertility, predicted floral traits (Morgan 1992).

Gender-SpecificPhenotypic Selection PredictedResponse: Gender-Specific Contributions Theory predicts that sexual dimorphism results from di- The relativeweight of gender-specificcontributions to to- vergent selection via male and female fertility (reviewed tal selection response differed between females and her- Limits on Sexual Dimorphism S13

Gender-specificComponents of the Predicted Responseto Selection ForHermaphrodites ForFemales a a F GHF HHF Hm FHGF Flowerdate FallLeaf # FallLeaf size

SprLeaf# SprLeaf size Trichome# Runner# -2 0 2 4 6 8 -2 0 2 4 6 8 -4 Standarddeviation units (x 00)

Figure 2: Gender-specificcontributions to the total standardizedpredicted response to selection in hermaphroditeand female Fragariavirginiana. Gender-specificcontributions are estimated as indicated in equation (2). maphrodites. Trait means for females of the next gener- is not expectedto change and those traitsfor which change ation are determined primarily by the female-specific would be expected if all else remained constant. Little or contribution to total response (fig. 2) because of a pre- no change in sexual dimorphismwas predictedfor fall leaf ponderance of indirect components that facilitate rather traits, runner number, and flowering date, and this is in than retarddirect response to selection (table 3). In con- line with the lack of sexual dimorphism in those traits trast, the male-specific contribution to female total re- when measured in F. virginiana'sdioecious sister species sponse is small because of weakerbetween-sex covariance, Fragariachiloensis (Hancock and Bringhurst1980). On the which is furtherdiminished by indirectresponses opposing other hand, calculationspresented here predict greaterdi- the direct responses for almost all traits. Trait means for morphism in spring leaf size and number (females are hermaphrodites of the next generation are determined predicted to produce fewer, smaller leaves than hermaph- slightly more equitably by the three gender-specificcon- rodites). At first glance, this prediction seems contraryto tributions. Not only does each gender-specificcontribu- the patternseen in many dioecious species;that is, females tion have a more equal number of antagonistic and fa- usually have larger leaves and fewer branches (table 1). cilitating indirect responses contributingto total response However, this trend is dominated by woody shrubs and (table 3), but response to selection via female fertility of in particularby a single genus (18 species of Leucaden- hermaphroditescontributes a large portion of the total dron).Even among the three herbaceousdioecious species, response. These results emphasize the importance of un- F. chiloensisis the only one to show a trend for smaller derstanding both between- and within-sex genetic co- leaves in females than in males (~10% smaller; P = variation (e.g., Meagher 1992;Costich and Meagher2001; 0.12; Hancock and Bringhurst1980). Thus, the predicted Ashman 2003; Carusoet al. 2003; Delph et al. 2004) before responses for several of the traits studied here in F. vir- drawing conclusions about the independent evolution of giniana are concordant with the direction and degree of traits in the sex morphs (see also Reeve and Fairbairn observed sexual dimorphism in its closest dioecious 1999). relative.

PredictedResponse: Sexual Dimorphism Considerationof UnderlyingAssumptions Predictedresponses in sexual dimorphism fell loosely into The predictions made here have several underlying as- two categories:those traits for which sexual dimorphism sumptions. The first assumption is that the genetic vari- S14 The AmericanNaturalist

Predictedchange in trait mean 6.00

- / Flowerdate 4.20 - CC 4 FallLeaf # FallLeaf size 2.40 2' 4 I Spr Leaf# 0 S 0.60 ? SprLeafsize + Trichome# - . S -1.20 , V Runner# 7 /+ -3.00 -3.00 -1.20 0.60 2.40 4.20 6.00

infemales (1 OOsdu)

Figure 3: Predicted change in sexual dimorphism after one generation of selection on vegetative and phenological traits of Fragariavirginiana in a field garden. Predicted responses are shown in hundreds of standarddeviation units and were calculatedaccording to equations (1) and (2). The dashed line reflects no predicted change in sexual dimorphism.

ance-covariancematrix (G) remains constant over time Conclusions (reviewedin Steppanet al. 2002). A study on reproductive an of sexual traits in this species was not able to detect population By providing integrated study dimorphism, this work sheds first on the of modest di- differences in G for reproductivetraits (Ashman 2003), light genesis in traitsin This and if spatial variation can be viewed as a surrogatefor morphism vegetative gynodioeciousplants. work revealed that sometimes differ- temporal variation, then this suggests that assuming that selection, although ential, was male and female fer- G is relatively constant is reasonable here. Second, pre- rarelydivergent through dictions made here assume that the traits are not under tility modes. In addition, estimation and inspection of to selection that low conflicting selection pressures at other life stages or via predicted responses suggest genetic variationand covariationlimit other of fitness 1991; G6mez pervasivegenetic response. components (e.g., Campbell these data that limited sexual dimor- 2004) or under indirect selection via correlatedtraits not Together, suggest phism is due to genetic covariation between traits within considered here. This latter assumption may be violated sex morphs modifying similar selection rather than solely under natural pollination conditions because petal size is to strong between sex-genetic covariance restricting di- positively correlatedwith spring leaf size (T.-L. Ashman, vergent selection. It also demonstrates that complete in- unpublished data). If selection via pollinator preferences formation on genetic covariation, direct and indirect se- favors increasedpetal size (e.g., Ashman and Diefenderfer lection, and genetic contributions to the next generation 2001; A. L. Case and T.-L. Ashman, unpublished manu- is needed before one can fully evaluate the evolutionary it could facilitate total to selection in script), response dynamics of sexual dimorphism. hermaphroditesbut retardresponse in females.Pollinator- mediated selection is assumed not to have played a role here because all plants were hand pollinated,but this sce- Acknowledgments nario could explainthe discordancebetween predicted sex- Thanks to S. Warnagirisfor implementing the crossing ual dimorphism in spring leaf size in F. virginianaand design; E. Yorkfor greenhouse support; J. Floyd, N. Grey, that observed in the wild. A final assumption is that es- S. Gryger, L. Hernandez, S. Kalla, J. Thompson, S. Zel- timates of fertility used here are reasonablesurrogates for gowski, and D. Zyger for scoring assistance;the staff of actual fertility. Pymatuning Laboratoryof Ecology and C. Martin at the Limits on SexualDimorphism S15

PennsylvaniaFish Hatcheryfor logisticsupport; J. Lawrence and whole-plant gender on floral evolution in Ecballium elatrium for writingthe resamplingprograms; L. Delph, D. Fairbairn, (Cucurbitaceae).Biological Journal of the LinneanSociety 74:475- 487. and an anonymous reviewerfor comments on the manu- Dawson, T. E., and M. A. Geber. 1999. Sexual in script.This researchwas supportedby the NationalScience dimorphism physiology and morphology. Pages 175-215 in M. A. Geber, T. E. Daw- Foundation(grants DEB-9707247, DEB-9903802, and DEB- son, and L. E Delph, eds. Gender and sexual dimorphism in flow- 0109099 and REU This is contribution 162 supplements). ering plants. Springer, Berlin. to the PymatuningLaboratory of Ecology. Delph, L. E 1999. Sexual dimorphism in life history. Pages 149-174 in M. A. Geber, T. E. Dawson, and L. E Delph, eds. Gender and sexual in Berlin. LiteratureCited dimorphism flowering plants. Springer, Delph, L. E, E M. Frey, J. C. Steven, and J. L. Gehring. 2004. In- the evolution of the size of floral Ahmadi, H., and R. S. Bringhurst. 1989. Genetics of sex expression vestigating independent organs via G-matrix estimation and artificial selection. Evolution and De- in Fragaria species. American Journal of Botany 78:504-514. 6: 438-448. Alonso, C., and C. M. Herrera. 2001. Neither vegetative nor repro- velopment L. 1967. Sex in Israel for Scientific ductive advantages account for high frequency of male-steriles in Dzhaparidze, I. plants. I. Program southern Spanish gynodioecious Daphne laureola (Thymelae- Translations, Jerusalem. aceae). American Journal of Botany 88:1016-1024. Eckhart, V. M. 1999. Sexual dimorphism in flowers and inflores- Ashman, T.-L. 1999. Determinants of sex allocation in a gynodioe- cences. Pages 123-148 in M. A. Geber, T. E. Dawson, and L. E cious wild strawberry:implications for the evolution of dioecy and Delph, eds. Gender and sexual dimorphism in flowering plants. sexual dimorphism. Journal of Evolutionary Biology 12:648-661. Springer, Berlin. . 2003. Constraints on the evolution of dioecy and sexual Eckhart, V. M., and F. S. I. Chapin. 1997. Nutrient sensitivity of the dimorphism:field estimatesof quantitativegenetic parameters for cost of male function in gynodioecious Phacelia linearis (Hydro- reproductive traits in three populations of gynodioecious Fragaria phyllaceae). American Journal of Botany 84:1092-1098. virginiana. Evolution 57:2012-2025. Elle, E., and T. R. Meagher. 2000. Sex allocation and reproductive Ashman, T.-L., and C. Diefenderfer. 2001. Sex ratio represents a success in the andromonecious perennial Solanum carolinense unique context for selection on attractive traits: consequences for (Solanaceae). II. Paternity and functional gender. American Nat- the evolution of sexual dimorphism. American Naturalist 157:334- uralist 156:622-636. 347. Etterson, J. R., and R. G. Shaw. 2001. Constraint to adaptive evolution B&gin,M., and D. A. Roff. 2001. An analysis of G matrix variation in response to global warming. Science 294:151-154. in two closely related cricket species, Gryllusfirmus and G. penn- Geber, M. A. 1999. Theories of the evolution of sexual dimorphism. sylvanicus. Journal of Evolutionary Biology 14:1-13. Pages 97-122 in M. A. Geber, T. E. Dawson, and L. E Delph, eds. Bond, W. J., and K. E. Maze. 1999. Survival costs and reproductive Gender and sexual dimorphism in flowering plants. Springer, benefitsof floraldisplay in a sexuallydimorphic dioecious shrub, Berlin. Leucadendronxanthoconus. Evolutionary Ecology 13:1-18. Geber, M. A., and L. R. Griffen. 2003. Inheritance and natural se- Bond, W. J., and J. Midgley. 1988. Allometry and sexual differences lection on functional traits. International Journal of Plant Sciences in leaf size. American Naturalist 131:901-910. 164(suppl.):S21-S42. D. R. 1991. Effects of floral traits on Campbell, sequential compo- G6mez, J. M. 2004. Bigger is not always better: conflicting selective nents of fitness in American Naturalist 137: Ipomopsis aggregata. pressures on seed size in Quercus ilex. Evolution 58:71-80. 713-737. Hancock, J. E, Jr., and R. S. Bringhurst. 1980. Sexual dimorphism --. 1996. Evolution of floral traits in a hermaphroditic plant: in the strawberry Fragaria chiloensis. Evolution 34:762-768. field measurements of heritabilities and genetic correlations. Evo- Kohn, J. R. 1989. Sex ratio, seed production, biomass allocation, and lution 50:1442-1453. the cost of male function in Cucurbitafoetidissima HBK (Cucur- Caruso, C. M. 2004. The quantitative genetics of floral trait variation bitaceae). Evolution 43:1424-1434. in Lobelia: potential constraints on adaptive evolution. Evolution Kohorn, L. U. 1994. Shoot and in 58:732-740. morphology reproduction jojoba: advantages of sexual dimorphism. Ecology 75:2384-2394. Caruso, C. M., H. Maherali, and R. B. Jackson. 2003. Gender-specific . 1995. Geographic variation in the occurrence and extent of floral and physiological traits: implications for the maintenance sexual dimorphism in a dioecious shrub, Simmondsia chinensis. of females in gynodioecious Lobelia siphilitica. Oecologia (Berlin) Oikos 74:137-145. 135:524-531. H. 1992. Patterns of resource allocation in male and Case, A. L., and T.-L. Ashman. 2005. Gender dimorphism: impli- Korpelainen, female of Rumex acetosa and Rumex acetosella. cations for reproductive cost and gender-specific physiology. Pages plants Oecologia 126-154 in E. Reekie and E Bazzaz, eds. The allocation of resources (Berlin) 89:133-139. and S. Arnold. 1983. The measurement of selection on to reproduction in plants. Elsevier, New York. Lande, R., J. correlated characters. Evolution 37:1210-1226. Cipollini, M. L., and D. E Whigham. 1994. Sexual dimorphism and cost of reproduction in the dioecious shrub Lindera benzoin (Lau- Lovett Doust, J. L., and L. L. Lovett Doust. 1988. Modules of pro- raceae). American Journal of Botany 81:65-75. duction and reproduction in a dioecious clonal shrub, Rhus ty- Conner, J., and S. Via. 1992. Natural selection on body size in Tri- phina. Ecology 69:741-750. bolium: possible genetic constraints on adaptive evolution. Hered- Lovett Doust, J., G. O'Brien, and L. Lovett Doust. 1987. Effect of ity 69:73-83. density on secondary sex characteristicsand sex ratio in Silene alba Costich, D. E., and T. R. Meagher. 2001. Impacts of floral gender (Caryophyllaceae). American Journal of Botany 74:40-46. S16 The AmericanNaturalist

Lynch, M., and B. Walsh. 1998. Genetics and analysis of quantitative Ramsey, M., and G. Vaughton. 2001. Sex expression and sexual di- traits. Sinauer, Sunderland, MA. morphism in subdioecious Wurmbeadioica (Colchicaceae). Inter- Lyons, E. E., D. Miller, and T. R. Meagher. 1994. Evolutionary dy- national Journal of Plant Sciences 162:589-597. namics of sex ratio and gender dimorphism in Silene latifolia. I. -- . 2002. Maintenance of gynodioecy in Wurmbeabiglandulosa Environmental effects. Journal of Heredity 85:196-203. (Cholicaceae): gender differences in seed production and progeny Machon, N., V. D. Le Boulc'h, and C. Rameau. 1995. Quantitative success. Plant Systematics and Evolution 232:189-200. analysis of sexual dimorphism in Asparagus.Canadian Journal of Reeve, J. P., and D. J. Fairbairn. 1999. Change in sexual size dimor- Botany 73:1780-1786. phism as a correlated response to selection on fecundity. Heredity Meagher, T. R. 1992. The quantitative genetics of sexual dimorphism 83:697-706. in Silene latifolia (Caryophyllaceae). I. Genetic variation. Evolution Roff, D. A. 1997. Evolutionary quantitative genetics. Chapman & 46:445-457. Hall, New York. . 1999. The quantitative genetics of sexual dimorphism. Pages Sakai, A. K., and T. A. Burris. 1985. Growth in male and female 275-290 in M. A. Geber, T. E. Dawson, and L. F. Delph, eds. Gender aspen (Populus tremuloides)clones: a twenty-five-year longitudinal and sexual dimorphism in flowering plants. Springer, Berlin. study. Ecology 66:1921-1927. Mitchell, J. R., R. G. Shaw, and N. M. Waser. 1998. Pollinator se- Sakai, A. K., and T. L. Sharik. 1988. Clonal growth of male and lection, quantitative genetics and predicted evolutionary responses female bigtooth aspen Populus grandidentata. Ecology 69:2031- of floral traits in Penstemon centranthifolius (Scrophulariaceae). 2033. International Journal of Plant Sciences 159:331-337. Shykoff, J. A., S.-O. Kolokotronis, C. L. Collin, and M. L6pez-Villa- Morgan, M. T. 1992. The evolution of traits influencing male and vicencio. 2003. Effects of male sterility on reproductive traits in female fertility in outcrossing plants. American Naturalist 139: gynodioecious plants: a meta-analysis. Oecologia (Berlin) 135:1- 1022-1051. 9. Morgan, M. T., and T.-L. Ashman. 2003. Quantitative character evo- Staudt, G. 1989. The species of Fragaria, their taxonomic and geo- lution under complicated sexual systems, illustrated in gynodioe- graphical distribution. Acta Horticulturae 265:23-33. cious Fragaria virginiana. American Naturalist 162:257-264. Steppan, S. J., P. C. Phillips, and D. Houle. 2002. Comparative quan- Nicotra, A. B. 1999. Sexually dimorphic growth in the dioecious titative genetics: evolution of the G matrix. Trends in Ecology & tropical shrub, Siparuna grandiflora. Functional Ecology 13:322- Evolution 17:320-327. 331. Worley, A. C., and S. C. H. Barrett. 2000. Evolution of floral display Olff, H., D. Kuiper, J. M. M. Van Damme, and P. J. C. Kuiper. 1989. in Eichorniapaniculata (Pontederiaceae): direct and correlated re- Gynodioecy in Plantago laneolata. VI. Functions of cytokinins in sponses to selection on flower size and number. Evolution 54: growth, development and reproduction of two sex types. Canadian 1533-1545. Journal of Botany 67:2765-2769. Poot, P. 1997. Reproductive allocation and resource compensation in male-sterile and hermaphroditic plants of Plantago lanceolata (Plantaginaceae). American Journal of Botany 84:1256-1265. Symposium Editor:Lynda Delph